GalenOps prod · 0.1.0

Agent

Data cleaned

Horizon 1 cdm_agent

18 edit checks swept AE, DM, SV, EX and LB; deterministic queries auto-closed.

Escalates: Queries open longer than the 14-day SLA, and any query an automated edit check cannot resolve, as medium-priority tasks — to the data manager.

Agents

Runs

3

Escalations

2

Failures

0

Avg duration

4.1 s

Literature & benchmarks

  • JSCDM (Amatya et al. 2021) — 14-day query-resolution SLA as the aging threshold
  • Edit-check taxonomy — range, consistency and missing-value checks generated from the CRF

Run history (3 across all studies)

Newest first, 50 per page. Expand a row for the recorded reasoning and evidence.

Expand When Study Outcome Model Duration Reasoning
2026-09-29 23:42 UTC A Multicenter, Randomized, Open-Label Phase 2b Study to Investigate the Preliminary Efficacy and Safety of INNO-206 (Doxorubicin-EMCH) Compared to Doxorubicin in Subjects With Metastatic, Locally Advanced, or Unresectable Soft Tissue Sarcoma escalated human claude-sonnet-5 3.5 s The CDM sweep of this Phase 2b INNO-206 vs doxorubicin sarcoma study (140 subjects) ran 18 edit checks; 2 fired: AE-006 "Grade ≥4 AE not flagged serious" (16 hits, P1) and VS-001 "Visit outside protocol window" (70 hits, P3), totaling 86 discrepancies. All 86 were routed to a human with 0 auto-closed and 0 deduplicated, and machine-answered share is 0.0% versus the Amatya & Edgerton 2021 (JSCDM) query accuracy target of ≥90%. All 86 queries are past the 14-day SLA, and all 86 are also past 30 days, with 16 of them P1 priority — exceeding the JSCDM listing-review benchmark of ≥80% within two weeks (target 95%). A review task has been opened. The human reviewer should decide how to prioritize and resolve the 16 overdue P1 AE-006 seriousness discrepancies versus the 70 VS-001 visit-window queries.
2026-09-29 21:13 UTC A Phase III, Randomized, Double-Blind Study of Torivumab Plus Chemotherapy Versus Placebo Plus Chemotherapy in Previously Untreated Advanced Non-Small Cell Lung Cancer completed claude-sonnet-5 4.6 s A CDM sweep of this Phase III NSCLC trial (317 subjects) ran 18 edit checks, with 3 firing: AE-001 "AE onset before first dose" (4 hits, P2), AE-005 "Serious AE without seriousness criteria" (1 hit, P2), and AE-007 "Duplicate AE record" (2 hits, P3, deterministic). This produced 7 deduplicated discrepancies, all 7 already queried, with 0 new queries generated and 0 auto-closed or left open for a data manager. Aging shows 4 open queries past the 14-day SLA and the same 4 past 30 days, though 0 of these are P1 priority. Per the cited benchmark (Amatya & Edgerton 2021, JSCDM), listing review target is ≥80% within 2 weeks (target 95%) and query accuracy ≥90%; query_accuracy_proxy is not available here. No review task was opened and nothing was routed to a human in this sweep. A human should decide whether the 4 queries aging past 14/30 days require expedited follow-up given the benchmark targets.
2026-09-19 08:58 UTC A Phase III, Randomized, Double-Blind Study of Torivumab Plus Chemotherapy Versus Placebo Plus Chemotherapy in Previously Untreated Advanced Non-Small Cell Lung Cancer escalated human claude-sonnet-5 4.3 s The CDM sweep of this Phase III NSCLC study (317 subjects) ran 18 edit checks; 3 fired: AE-001 (AE onset before first dose) ×4, AE-005 (serious AE without seriousness criteria) ×1, and AE-007 (duplicate AE record, deterministic) ×2. Of 7 discrepancies generated, 2 were auto-closed and 5 remain open, routed to a human. Machine-answered accuracy is 29% (0.286), well below the JSCDM benchmark target of ≥90% query accuracy. Aging shows 4 open queries past the 14-day SLA and 4 past 30 days, though 0 are P1 priority. Samples include AE onset-before-dose cases (e.g., SITE010-0221) and one serious AE (SITE001-0066) missing seriousness criteria. A review task has been opened. The human reviewer is asked to decide how to re-prioritize and resolve the 5 open queries, including the overdue ones and the undocumented serious AE.